The application domain of crime investigation and prevention deals with incomplete information from a broad variety of data sources and a complex structure of interdependencies of rapidly changing situations. The paper describes an approach of integration of multiple statistical analyses for recognition of crime information patterns in data warehouse. With multiple statistical methods we develop a decision support system based on real data of warehouses of socialeconomic and crime indicators. The proposed multiple regression models allows gaining new insights into the structure of problem and developing strategies for crime prevention measures.